Computers & Industrial Engineering 76 (2014) 292–306

Contents lists available at ScienceDirect

Computers & Industrial Engineering

journal homepage: www.elsevier.com/locate/caie

The impact of information sharing and inventory control coordination on

supply chain performances q
Francesco Costantino a, Giulio Di Gravio a, Ahmed Shaban a,b,⇑, Massimo Tronci a
a
Department of Mechanical and Aerospace Engineering, University of Rome ‘‘La Sapienza’’, Via Eudossiana, 18, 00184 Roma, Italy
b
Department of Industrial Engineering, Faculty of Engineering, Fayoum University, 63514 Fayoum, Egypt

a r t i c l e i n f o a b s t r a c t

Article history: The lack of coordination in supply chains can cause various inefficiencies like bullwhip effect and inven-
Received 16 October 2013 tory instability. Extensive researches quantified the value of sharing and forecasting of customer demand,
Received in revised form 2 August 2014 considering that all the supply chain partners can have access to the same information. However, only
Accepted 12 August 2014
few studies devoted to identify the value of limited collaboration or information visibility, considering
Available online 22 August 2014
their impact on the overall supply chain performances for local and global service level. This paper
attempts to fill this gap by investigating the interaction of collaboration and coordination in a four-
Keywords:
echelon supply chain under different scenarios of information sharing level. The results of the simulation
Supply chain
Information sharing
study show to what extent the bullwhip effect and the inventory variance increase and amplify when a
Bullwhip effect periodic review order-up-to level policy applies, noting that more benefits generate when coordination
Inventory variance starts at downstream echelons. A factorial design confirmed the importance of information sharing
Service level and quantified its interactions with inventory control parameters, proving that a poor forecasting and
Simulation definition of safety stock levels have a significant contribution to the instability across the chain. These
results provide useful implications for supply chain managers on how to control and drive supply chain
performances.
 2014 Elsevier Ltd. All rights reserved.

1. Introduction
Similar to the shape of a cowboy cracking his lash, the bullwhip
effect (Fig. 1) is the amplification of the demand variance as infor-
mation flows across the supply chain, from customer to factory
(Chatfield, Kim, Harrison, & Hayya, 2004; Lee, Padmanabhan, &
Whang, 1997b). Starting from the most famous experiences at
Campbell Soup’s (Fisher, Hammond, Obermeyer, & Raman, 1997),
HP and Procter & Gamble (Lee, Padmanabhan, & Whang, 1997a), a
clothing supply chain (Disney & Towill, 2003a), Glosuch (McCullen
& Towill, 2000) till the recent examples of fast moving consumer
goods (Zotteri, 2012) and car manufacturing (Klug, 2013), the bull-
whip effect continues to show its impact on supply chain perfor-
mances. Many researchers illustrated the bullwhip dynamics and
its negative effects: the amplification of order variance requires a
managerial effort that results in a consequent increase of costs to
q
This manuscript was processed by Area Editor Mohamad Y. Jaber.
⇑ Corresponding author at: Department of Mechanical and Aerospace Engineer-
ing, University of Rome ‘‘La Sapienza’’, Via Eudossiana, 18, 00184 Roma, Italy. Tel.:
+39 3282777914; fax: +39 0644585746.
E-mail addresses: francesco.costantino@uniroma1.it (F. Costantino), giulio.
digravio@uniroma1.it (G. Di Gravio), ahmed.shaban@uniroma1.it, ahmed.shaban@
fayoum.edu.eg (A. Shaban), massimo.tronci@uniroma1.it (M. Tronci).
maintain a target service level (Chatfield et al., 2004) and a further
reduction of business performances due to:
- High levels of inventory to face unexpected variations of the
demand with a relative increase of stocking costs, as the
amount of safety stocks derives from the order variability.
- Low service level to customers for unexpected stock-outs that,
in the worst cases, can cause the cancelation of orders and the
reduction of actual and future sales.
- Reduction of quality for the necessity to increase production
rates to satisfy peaks of demand.
- Increase of costs for rescheduling, updating of production plans
and reworking.
Besides the role of misperception of feedback by managers and
of operational causes, accurately described starting from the
research of Sterman (1989), there is still a substantial need to
understand the contribution that coordination and collaboration
could give to counteract it in a modern and competitive
environment.
According to these considerations, this paper presents the fol-
lowing structure: (1) a literature review in Section 2 provides the
motivations and the scope of the research, determining the specific

http://dx.doi.org/10.1016/j.cie.2014.08.006
0360-8352/ 2014 Elsevier Ltd. All rights reserved.
 F. Costantino et al. / Computers & Industrial Engineering 76 (2014) 292–306
Fig. 1. An example of demand amplification in a linear supply chain.
field of analysis; (2) Section 3 describes the supply chain simula-
tion model and the performance measurement system; (3)
Section 4 provides a quantification of the impact of different infor-
mation sharing levels on supply chain performances; (4) Section 5
shows the interactions of the inventory control parameters on the
performances, according to different levels of information sharing;
(5) Section 6 gives general insights and implications for managers,
and Section 7 summarizes the conclusions.
2. Literature review
Forrester (1958) was the first to observe and describe the bull-
whip effect, calling it ‘‘demand amplification’’, determining that it
is due to natural system dynamics and delays in supply chain.
Reducing the lead-time between echelons and adopting coordina-
tion approaches such as vendor-managed inventory, to eliminate
decision-making layers and increase visibility on information
without distortion, can help to smooth this effect (Disney &
Towill, 2003b; Lee et al., 1997b).
Two classes of sources cause the bullwhip effect: behavioral
causes (Nienhaus, Ziegenbein, & Schoensleben, 2006; Rong, Shen,
& Snyder, 2008; Sterman, 1989) and operational causes (Lee
et al., 1997a, 1997b). The first refers to the evidence that decision
makers consistently underweight the supply line when making
order decisions (Croson & Donohue, 2006) while the second relates
to the operational conditions of the replenishment process as for
demand signal processing, non-zero lead-times, order batching,
price fluctuations, rationing and shortage gaming.
In particular, demand signal processing encompasses the fore-
casting of future demand and inventory adjustment through apply-
ing inventory control policies, proven by Disney and Lambrecht
(2008) to be of a significant importance. In this stream of research,
many studies attempted to identify the impact on the bullwhip
effect of different inventory control policies, evaluating the role of
forecasting methods (Barlas & Gunduz, 2011; Chen, Drezner, Ryan,
& Simchi-Levi, 2000; Chen, Ryan, & Simchi-Levi, 2000;
Dejonckheere, Disney, Lambrecht, & Towill, 2003, 2004; Hosoda &
Disney, 2006a, 2006b; Wright & Yuan, 2008; Zhang, 2004), lead-
times (Chatfield, 2013; Chatfield et al., 2004; Chen, Drezner, et al.,
2000; Chen, Ryan, et al., 2000; Ciancimino, Cannella, Bruccoleri, &
Framinan, 2012; Dejonckheere et al., 2003, 2004; Kelepouris,
Miliotis, & Pramatari, 2008) and information sharing (Chatfield &
Pritchard, 2013; Chatfield et al., 2004; Chen, Drezner, et al., 2000;
Dejonckheere et al., 2004; Kelepouris et al., 2008) to give useful
insights for supply chain managers on how to handle demand vari-
ability amplification. A specific attention is to give to side effects of
bullwhip effect, as for inventory stability, especially in multi-
echelon supply chains: while the amplification of the demand con-
tributes to the increase of the production and inventory costs, the
293
variance of the net inventory determines the ability of any single
organization to meet a service level in a cost-effective manner
(Hussain, Shome, & Lee, 2012; Ma, Wang, Che, Huang, & Xu, 2013).
To this extent, most studies focused on the periodic review order-
up-to policy because of its popularity (Chatfield & Pritchard, 2013;
Dejonckheere et al., 2003; Disney & Lambrecht, 2008; Wright &
Yuan, 2008). The order-up-to policy is a common strategy when
there is no fixed ordering cost (e.g. in case of visiting agents or
framework contracts with suppliers) or both holding and shortage
costs are proportional to the volume of the on-hand inventory
(Dejonckheere et al., 2003). The important role of this inventory
control rule in many supply chains derives from the small set of
parameters to set as the only information for its implementation
are a forecasting value of demand and a level of safety stock, to
determine according to a required service level. Of particular inter-
est, Chen, Drezner, et al. (2000) quantified the bullwhip effect in a
two-echelon serial supply chain experiences auto-regressive AR(1)
demand process and employs the order-up-to level with the moving
average method, and they further extended the results for a multi-
echelon supply chain both with and without information sharing.
Chen, Ryan, et al. (2000) extended the analysis for a supply chain
employs the order-up-to level with the exponential smoothing
method. Xu, Dong, and Evers (2001) obtained similar results using
the exponential smoothing technique. Chen, Ryan, et al. (2000)
showed also that, if the forecasting parameters for both exponential
smoothing and moving average methods are set to achieve the same
forecasting accuracy, then moving average gives a lower order var-
iance. Dejonckheere et al. (2004) confirmed these results using a
control theoretic approach. Chatfield et al. (2004) examined the
effects of stochastic lead times, information sharing and quality of
information in a periodic review order-up-to level inventory system
through a simulation study. They considered three measures to
track the supply chain performance, namely, standard deviation of
the orders at each echelon, total variance amplification (i.e. bull-
whip effect) and stage variance amplification. They concluded that
lead-time variability exacerbates variance amplification and that
information sharing and information quality are highly significant
to handle the propagation of variability. Jakšič and Rusjan (2008),
Kelepouris et al. (2008), Wright and Yuan (2008), Hussain et al.
(2012), Chatfield (2013) and Chatfield, Hayya, and Cook (2013) con-
ducted similar studies to confirm these findings in more complex
configurations of supply chains or different inventory control
conditions.
Different mitigation strategies can help to overcome the effect
of demand signal processing, from improving the operational effi-
ciency to the use of collaboration approaches in decision making
(Campuzano-Bolarín, Mula, & Peidro, 2013; Costantino, Di Gravio,
Shaban, & Tronci, 2013a; Costantino, Di Gravio, Shaban, & Tronci,
2013b; Dejonckheere et al., 2004; O’Donnell et al., 2006). In
294 F. Costantino et al. / Computers & Industrial Engineering 76 (2014) 292–306

particular, Holweg, Disney, Holmström, and Småros (2005) classi- echelons employ a periodic review order-up-to (R, S) inventory
fied supply chain collaboration initiatives on inventory replenish- policy in which the order-up-to level updates at the end of each
ment and forecasting dimensions, concluding that any action review period (R), with R = 1, according to the forecasting of the
have to integrate both these areas to reach significant improve- future demand.
ments. Interventions that improve, locally or globally, the level of The model presents the following assumptions (Ciancimino
coordination showed their effectiveness in practice (Chatfield et al., 2012; Sterman, 1989; Wright & Yuan, 2008):
et al., 2004; Cho & Lee, 2013; Costantino et al., 2013a, 2013b;
Kelepouris et al., 2008; Ye & Wang, 2013).  The factory has an unlimited capacity to produce any quantity
Most of previous works assumed that all partners use the same ordered by the distributor.
forecasting methods and inventory control policy with the same  The stocking capacity at any echelon is unlimited.
parameters. In addition, they assumed that, if there is collaboration  The unfulfilled orders, due to out of stock, at any echelon are not
(i.e. information sharing of customer demand), all partners partic- lost but they become backlogs, to satisfy as soon as the inven-
ipate. The paper extends these efforts through measuring perfor- tory recovers.
mances under different levels of information sharing, forecasting  The transportation capacity between adjacent echelons is
parameters and safety stock level at the upstream and downstream unlimited.
echelons of the supply chain. This results in different scenarios of  The ordering and delivery lead-times are deterministic and
partial collaboration to quantify the effectiveness to what extent fixed across the supply chain with ordering lead-time (Lo ¼ R)
the coordination of policies among partners can help in smoothing = 1 and delivery lead-time (Ld) = 2.
the bullwhip effect in a multi-echelon supply chain. To allow  Orders are always positive or equal to zero, cancelations are not
comparisons with the main cited research, the partners apply a allowed (non-negativity condition).
periodic review order-up-to policy with a moving average forecast-
ing method where the amplification of the demand, the inventory The Eqs. (1)–(7) define the different state variables at each
variance and the service level are measures of performance, coher- echelon i in each period t; where i = 1, ..., 4; e.g., i = 1 stands for
ently to Kurien and Qureshi (2011) review and Tangen (2004) the retailer and i = 4 stands for the factory, and we also refer to
framework. To this extent and according to the classification of the customer as i = 0 to adjust the model:
Chatfield (2013), simulation is the most appropriate approach to
study supply chain complex dynamics. Similar experiences proved SLit ¼ SLit1 þ Oit1  SRiþ1
tLd ð1Þ
the effectiveness of discrete-event simulation (Tako & Robinson,
SRit ¼ MinfIOit þ Bit1 ; Iit1 þ SRiþ1
tLd g ð2Þ
2012; Chatfield, 2013; Costantino, Di Gravio, Shaban, & Tronci,
2014a, 2014b, 2014c; Costantino et al., 2013a; Lau, Xie, & Zhao,
IOit ¼ OtLo
i1
ð3Þ
2008). An experimental design evaluates the statistical significance
of the interactions of forecasting parameters, information sharing Iit ¼ Iit1 þ SRiþ1 i
tLd  SRt ð4Þ
level and safety stock level to evaluate how they can contribute
to improve or reduce performances. The analysis aims at giving Bit ¼ Bit1 þ IOit  SRit ð5Þ
some useful insights and managerial implications to determine
guidelines of coordination in a competitive environment. NIit ¼ Iit  Bit ð6Þ

IPit ¼ Iit þ SLit  Bit ð7Þ

3. Supply chain simulation modeling
Following the leading research of Chen, Drezner, et al. (2000),
Chatfield et al. (2004) and Dejonckheere et al. (2004), this study
propose a discrete event simulation to identify the role of the
ordering policy in controlling bullwhip effect, where the perfor-
mances at each echelon derives mainly from the level of coordina-
tion among the partners of the supply chain.
The model represents a traditional four-echelon supply chain
In each period t, each echelon i receives an amount of shipment
SRiþ1
tLd issued by the upstream echelon i + 1 at time t  Ld. Thus,
the initial inventory level Iit1 increases by the shipment SRiþ1
tLd
and at the same time decreases by the amount of SRit released down
to the echelon i  1 (see, Eq. (4)). The amount SRit to ship to echelon
i  1 is the lower value between the initial inventory Iit1 added to
the incoming shipment SRiþ1 i
tLd and the order IOt added to the backlog

consisting of a customer, a retailer, a wholesaler, a distributor
and a factory (see, Fig. 2) as for many previous studies in this field
(Chatfield et al., 2004; Chatfield, 2013; Ciancimino et al., 2012;
Costantino et al., 2013a, 2013b, 2014a). Fig. 2 depicts a visual rep-
resentation of the supply chain in which, at any period, each ech-
elon i receives orders from its downstream partner i  1, satisfies
these orders from his own stock and then issues an order to
echelon i + 1. The retailer observes and satisfies the customer
demand Dt and places orders with the wholesaler. All the
order Bit1 (see, Eq. (2)) where at the retailer (echelon i = 1), the
incoming order at time t is the observed customer demand at this
time (IOi¼1 ¼ Dt ). Thus, the net inventory NIit in Eq. (6) is equal to
t
the difference between the available inventory level Iit and the back-
log Bit at time t. The inventory position (7) is the accumulation of the
amount in supply line SLit in Eq. (1) and the net inventory level in Eq.
(6). The order Oit of the echelon i to the echelon i + 1 depends on the
ordering policy.

Fig. 2. A multi-echelon supply chain.

 F. Costantino et al. / Computers & Industrial Engineering 76 (2014) 292–306
The Eqs. (8)–(15) represent the rules of the order-up-to policy:
Oit ¼ MaxfðSit  IPit Þ; 0g ð8Þ
IP it ¼ Sit1  IOit ð9Þ
Sit ¼ LD^ i þ SSi
t t ð10Þ
SSit ¼ k ^ it
r ð11Þ
Sit ^ i þ kr
¼ LD ^ it ð12Þ
t r2D =lD
n
X i
^i ¼ 1
i
D t IO ; for 8i > 1 ð13Þ
ni j¼1 tjþ1
X n
^R ¼ 1
i
D Dtjþ1 ; for i ¼ 1
t
ni j¼1
ðat the retailerÞ ð14Þ
In the periodic review order-up-to policy, at the end of each
period t, each echelon issues a non-negative order Oit whenever
the inventory position IP it is lower than a specific target level Sit
(order-up-to level) as in Eq. (8). The target inventory position for
an echelon i at time t, Sit relies on the expected demand over the
total lead-time (LD ^ i ) where L = R + Ld and D
^ i is the one-period
t t
ahead demand forecast, as shown in Eq. (10), resulting in a
dynamic target level that, every period t, follows the updated
demand forecast. In order to account for demand variation, a safety
stock component adds to the equation of the target inventory posi-
tion, as in Eqs. (10)–(12). The amount of safety stock depends on
the variation of incoming orders during the lead-time period (L)
and the service level (k). Another common approach is to extend
the lead-time to calculate the safety stock, instead of depending
on the formula kr ^ it (Chatfield & Pritchard, 2013; Dejonckheere
et al., 2004). In this model, we set k = 0 and considered the safety
stock term by extending the lead-time period by Ki so that the tar-
get inventory position Sit becomes as follow:
Sit ¼ ðL þ K i ÞD
^i
t ð15Þ
The demand forecast (D ^ i ) is dynamically updated in each period t
t The average service level quantifies the percentage of items
with the moving average forecasting technique which is commonly
used in practice and research (Chen, Drezner, et al., 2000; Disney &
Lambrecht, 2008). The only parameter required for the moving
average forecasting is the number of past periods ni used to average
the demand. Specifically, at the end of each period t, all echelons
other than the retailer (echelon i > 1) estimate the expected demand
in the next period (one-period ahead demand forecast, D ^ i ) based on
t elon i, and IOit is the incoming order to echelon i. The effective sim-
the average of the incoming orders from the adjacent downstream
echelon over the most recent ni periods (IOtni þ1 ; . . . ; IOit ) (see,
i
Eq. (13)). Eq. (13) is a general equation that can be used to obtain
the demand forecast at any echelon where IOitjþ1 represents the
incoming order from the downstream echelon i  1 to echelon i at
time t  j + 1 where j = 1, ..., ni. At the retailer (echelon i = 1), the
demand forecast is adjusted in the same manner but based on the
actual customer demand data (Dtni þ1 ; . . . ; Dt ) as shown in Eq. (14)
which is a special case of (13). The mean demand over the lead time
is estimated by multiplying the next period’s demand forecast by
the lead time added to the safety stock parameter which determines
the target inventory position in Eq. (15).
The last Eq. (15) shows that each echelon i in the supply chain
can use his own forecasting method and his own parameter of the
ordering policy.
3.1. Performance measures
The study analyzes and quantifies supply chain under three dif-
ferent performances: bullwhip effect ratio, inventory variance ratio
and average service level.
295
3.1.1. Bullwhip effect ratio
The bullwhip effect ratio expresses the amplification of demand
variability across the supply chain. In particular, Chen, Drezner,
et al. (2000) quantified bullwhip effect (BWEi ) analytically in terms
of the variance of the orders (r2Oi ) at echelon i relative to the vari-
ance of the demand at the retailer:
r2Oi =lOi
BWEi ¼ ð16Þ
where lOi represents the average of the orders at echelon i, lD is the
average of the customer demand and r2D is the variance of the cus-
tomer demand. It is expected that lOi ¼ lD in the long term run and,
therefore, the measure turns into BWEi ¼ r2Oi =r2D . Dejonckheere
et al. (2004) used the same metric as a measure of bullwhip effect
and called it the ‘Variance Ratio’. They stated that Variance Ratio > 1
results in a bullwhip; Variance Ratio < 1 results in order smoothing;
Variance Ratio = 1 results in a ‘‘pass-on-orders’’ policy, where the
ordering pattern exactly follows the demand pattern.
3.1.2. Inventory variance ratio
The second measure is the inventory variance ratio. Disney and
Towill (2003a) were the first to propose this measure to evaluate
the degree of inventory stability, as it quantifies the fluctuations
in net inventory variability (r2NIi ) relative to the fluctuation in
demand variability (r2D ):
r2NIi
Inv Ri ¼ ð17Þ
r2D
It can also measure the amplification in inventory instability as we
move up the supply chain, similar to the bullwhip effect ratio. An
increase in inventory variance would result in higher holding and
backlog costs, lower service level and increasing average inventory
costs per period.
3.1.3. Average service level
delivered immediately by the echelon i to satisfy an incoming
i
order (Zipkin, 2000). Service level or fill rate (Slt ) computes every
review time R over the effective delivery time (i.e., IOit > 0) as in
Eq. (18). Its time series constitutes the history of the effectiveness
of the delivery system, where SRit stands for the shipment from
echelon i to echelon i  1, Bit1 stands for the initial backlog at ech-
ulation time is equivalent to the summation of all periods with
IOit > 0; hence, T eff 6 T .
8 i i
< SRt Bt1  100 if SRit  Bit1 > 0
i IOit
Slt ¼ ð18Þ
:
0 if SRit  Bit1 6 0
PT eff i
Sl
ASli ¼ t¼1 t ð19Þ
T eff
The average service level (ASli ) is computed only over the effective
simulation time (T eff ) as in Eq. (19).
3.2. Verification and validation of the simulation model
The simulation model, developed in SIMUL8, followed a verifi-
cation and validation process with the analytical work of Chen,
Drezner, et al. (2000), the control engineering work of
Dejonckheere et al. (2004) and the simulation work of Chatfield
et al. (2004). Chen, Drezner, et al. (2000) developed a closed-form
expression for quantifying the bullwhip effect ratio in a two-stage
supply chain, generalizing to multi-echelon systems through
296 F. Costantino et al. / Computers & Industrial Engineering 76 (2014) 292–306

Table 1
Simulation model validation results.

Comparison of TSC results L + Ki = 5"i ni = 19"i DN(100, 102)

Echelon i TSC (a) Chen et al. (2000) (b) Dejonckheere et al. (2004) (c) Chatfield et al. (2004) (d) % (a–b) % (a–c) % (a–d)
Retailer 1.67 1.66 1.67 1.67 0% 0.13% 0.12%
Wholesaler 3.00 2.77 2.99 2.99 8% 0.27% 0.32%
Distributor 5.74 4.61 5.72 5.72 20% 0.36% 0.36%
Factory 11.45 7.68 11.43 11.43 33% 0.16% 0.13%
Average percentage error (%) 15% 0.23% 0.23%

Table 2 and absolute precision level (half width/mean) lower than 5% on

Information sharing scenarios (incoming order to each echelon). the bullwhip effect measures (Chatfield et al., 2004; Chatfield
et al., 2013; Robinson, 2005).
The demand pattern follows a normal distribution with an aver-
age of 100 and a standard deviation of 10 (i.e., N(100, 102)). The mov-
ing average parameter was set to ni = 19, "i, and the safety stock
component was set to Ki = 2, i.e., the value of L + Ki = 5 where
L = R + Ld and R = 1 (Chatfield et al., 2004; Dejonckheere et al.,
2004). We applied these values in Eq. (20) to get the bullwhip effect
ratio at each echelon i based on Chen, Drezner, et al. (2000), where
i = 1, ..., m and m = 4, and Li = L + Ki to consider the safety stock com-
CD = customer demand, RO = retailer order, WO = wholesaler order, DO = distributor ponent. Table 1 shows that the simulation model works as expected.
order The closed form expression of Chen, Drezner, et al. (2000) does not
consider the interactions in the model (Chatfield, 2013, 2004),
therefore, there is a considerable difference between the average
multiple applications (i.e. a product of several individual stage
percentage error between (a, c, d) and (b). The average percentage
amplification values) as shown in the following equation:
error of 0.23% with the other two simulation models is mainly due
! to the different random seeds in the different models and to the
r2Oi Ym
2Li 2L2i non-negativity of orders (allowed in the original ones).
P 1þ þ 2 8i ð20Þ
rD i¼1
2 ni ni

Dejonckheere et al. (2004) considered a control theoretic approach
to quantify the bullwhip effect and validated their model with Chen,
Drezner, et al. (2000), whereas Chatfield et al. (2004) adopted a sim-
ulation approach to quantify the bullwhip effect and validated their
results with both Dejonckheere et al. (2004) and Chen, Drezner,
et al. (2000). We compare the results of our model with them all,
setting the same simulation and operational parameters in a valida-
tion test of 20 replications for a length of 5200 periods, each with a
warm-up period of 200 periods (Chatfield et al., 2004; Dejonckheere
et al., 2004). For the other simulation experiments in the paper,
since the simulation model is a non-terminating system, the num-
ber of replications are selected based on a 95% confidence level
4. The value of information sharing
In literature, a general assumption of information sharing is that
all the partners in the supply chain would have access in real time
to same information (Dejonckheere et al., 2004). Considering the
periodic review order-up-to level strategy, the only set of sensible
information is the value of the demand: having information on
inventory or capacity of the downstream and upstream echelons
wouldn’t change the value of the orders issued by any partner. In
this study, we are seeking to quantify the impact of different con-
figurations of partial demand sharing on supply chain perfor-
mances as for cases when downstream echelons might

Bullwhip effect ratio

40.0
35.0
30.0
25.0
BWE

20.0
15.0
10.0
5.0
0.0
Retailer Wholesaler Distributor Factory
Scenario # 1 2.14 5.19 13.62 34.89
Scenario # 2 2.14 2.45 6.01 15.83
Scenario # 3 2.14 2.45 2.75 6.80
Scenario # 4 2.14 2.45 2.75 3.05
Scenario # 5 2.14 5.19 6.00 6.79
Scenario # 6 2.14 5.19 13.62 15.84

Fig. 3. Bullwhip effect ratio in different information sharing scenarios.

 F. Costantino et al. / Computers & Industrial Engineering 76 (2014) 292–306 297

Inventory variance ratio

90.0
80.0
70.0
60.0

InvR
50.0
40.0
30.0
20.0
10.0
0.0
Retailer Wholesaler Distributor Factory
Scenario # 1 4.56 11.61 32.62 79.04
Scenario # 2 4.53 7.56 19.88 49.46
Scenario # 3 4.54 7.39 10.34 25.73
Scenario # 4 4.54 7.38 10.05 11.83
Scenario # 5 4.54 10.93 18.67 25.52
Scenario # 6 4.54 11.21 29.68 49.72

Fig. 4. Inventory variance ratio in different information sharing scenarios.

collaborate without coordinating with the upstream ones (Son &
Sheu, 2008). Table 2 illustrates a sample of significant information
sharing scenarios to represent a gradually increasing level of col-
laboration, where not all the partners have complete visibility on
the demand or orders:
1. Scenario #1 or Traditional Supply Chain (TSC): the benchmark
case described above, in which the retailer is the only partner
that has access to the customer demand while the other part-
ners receive the downstream echelon’s orders.
2. Scenario #2 or Downstream Collaboration (DC) both the retailer
and the wholesaler have access to the customer demand.
3. Scenario #3 all the partners in the supply chain, except the fac-
tory, have access to the customer demand.
4. Scenario #4 or Information-Enriched Supply Chain (IESC) highest
level of information sharing in which all the partners have
access to the customer demand.
5. Scenario #5 limited collaboration among the upstream echelons
where both the distributor and the factory have access to the
incoming orders from the retailer to the wholesaler, without
knowing the customer demand.
6. Scenario #6 or Upstream Collaboration (UC) very limited collabo-
ration between the distributor and the factory in which the fac-
tory has access to the incoming orders at the distributor.
The level of information sharing changes the input data of the
forecast at any echelon. In the scenarios #2, #3 and #4, all the part-
ners that have access to the demand use Eq. (14) instead of Eq.
(13), meaning that the target level is no more affected by the vari-
ations of the orders coming from the downstream echelons but it
relies directly on the customer demand. In the scenarios #5 and
#6, all the partners that have access to the downstream echelons
orders use these values in their forecasts, instead of their incoming
order time series, to represent the customer demand with a lower
distortion.
The simulation settings for the experiment are considered as
follows:
- customer demand follows the normal distribution with lD = 30
and r2D ¼ 32 where rD is selected significantly smaller than lD,
so that the occurrence of negative demand observations is
negligible;

Average service level

101.0

100.0

99.0
ASl%

98.0

97.0

96.0

95.0
Retailer Wholesaler Distributor Factory
Scenario # 1 100.0 99.9 98.7 95.6
Scenario # 2 100.0 100.0 99.7 98.0
Scenario # 3 100.0 100.0 100.0 99.6
Scenario # 4 100.0 100.0 100.0 100.0
Scenario # 5 100.0 100.0 99.9 99.7
Scenario # 6 100.0 100.0 99.3 98.2

Fig. 5. Average service level in different information sharing scenarios.

298 F. Costantino et al. / Computers & Industrial Engineering 76 (2014) 292–306
- all echelons in the supply chain use the same ordering policy
with Ki = 1, "i, and the same forecasting method with
ni = 10, "i where these settings are the average and common
values employed for these parameters in the related literature
(Chatfield et al., 2004; Dejonckheere et al., 2003;
Dejonckheere et al., 2004; Wright & Yuan, 2008). In particular,
when ni = 10, "i, the forecast variance is r2D^ ¼ r2D =ni ¼ 0:9 that
means a percentage error of 3% on the average value.
- for each scenario, the simulation model runs for 20 replications
of 1200 periods, each with a warm-up of 200 periods (95% con-
fidence level with absolute precision level lower than 5%).
Fig. 3 confirms that the bullwhip effect act in all the scenarios
(i.e., BWEi > 1) regardless information are shared or not. However,
Scenario #1 shows the highest bullwhip effect at all echelons with
a ratio geometrically increasing as moving from the retailer to the
factory and a variability about 35 times the customer demand. Sce-
nario #4 represents the lower bound of the bullwhip effect as the
distortion of the demand is null and the effect derives only from
the value of lead times and the structure of the inventory control
policy. Increasing the information sharing level among the
Fig. 6. The sensitivity of the bullwhip effect to the experimental factors.
downstream echelons shows a considerable effect throughout the
supply chain and especially at the most upstream echelons, as in
Scenarios #2–6. For instance, the collaboration between the retai-
ler and the wholesaler in Scenario #2 (downstream collaboration)
allows a decrease of more than 50% of the bullwhip effect, i.e. from
5.19 to 2.45 at the wholesaler, from 13.62 to 6.01 at the distributor
and from 34.89 to 15.83 at the factory. More collaboration leads to
a reduction in the propagation rate: bullwhip effect ratio increases
geometrically over the echelons that do not participate in collabo-
ration (e.g., Scenarios #1, #2, #6) and increase linearly over the
echelons that participate (e.g., Scenarios #3, #4, #5). Furthermore,
the results indicate that downstream collaboration is more effec-
tive than upstream collaboration as, once the amplification starts,
it is more difficult to counteract.
In Fig. 4, the behavior in terms of inventory variance ratio
resembles to a considerable degree the bullwhip effect in Fig. 3.
The lack of collaboration leads to a large increase in the inventory
instability moving upstream in the supply chain as a negative con-
sequence of the bullwhip effect. The collaboration through infor-
mation sharing have both local and global effect: comparing the
Upstream Collaboration (Scenario #6) with the Traditional Supply
 F. Costantino et al. / Computers & Industrial Engineering 76 (2014) 292–306 299

Fig. 7. Interaction effect between MA and Info_Shar on bullwhip effect.

Fig. 8. Interaction effect of the safety stock levels on the bullwhip effect.

Table 3
ANOVA results for the bullwhip effect at the different echelons.

Chain (Scenario #1), the inventory variance ratio decreases from The behavior of the average service level is a reverse pattern of
32.62 to 29.68 at the distributor (local effect), from 79.04 to the bullwhip effect and inventory variance, as for Fig. 5. It is clear
49.72 at the factory (global effect) and slightly from 11.61 to that the average service level (ASli) decreases as we move upstream
11.21 at the wholesaler (global effect). in the supply chain especially when the collaboration level is low
300 F. Costantino et al. / Computers & Industrial Engineering 76 (2014) 292–306

Fig. 9. The sensitivity of the inventory variance to the experimental factors.

Fig. 10. Impact of different settings of safety stock level on the inventory variance.

or absent. This conclusion confirms the recent study of Chatfield This means that the upstream echelons are not able to handle
et al. (2013) where they investigated the existence and magnitude properly the variability of the incoming orders although the down-
of stock out propagation and amplification in inventory systems. stream are successful to do that. The distributor and the factory
 F. Costantino et al. / Computers & Industrial Engineering 76 (2014) 292–306
that suffer the problem would have to increase the level of their
safety stock to handle this situation. Interestingly, as the level of
collaboration increases, the average service level improves at the
higher upstream echelons.
The best scenario that achieves the lowest bullwhip effect and
inventory variance ratio and the highest average service level is
the Information-Enriched Supply Chain in which all echelons have
access to the customer demand. The results show that, when pos-
sible, local partnerships (downstream or upstream) are anyway
recommended, as they can lead to decrease the variance amplifica-
tion at any echelon whilst achieving improvements on inventory
performance.
5. Experimental design
Forecasting and inventory control parameters are key causes of
the bullwhip effect and inventory instability that could influence
the impact of information sharing. To extend the results of the
analysis, we design a factorial experiment with five factors on
two levels each, according to the hypothesis of Chatfield et al.
(2004) and Barlas and Gunduz (2011):
1- Information sharing (Info_Shar): TSC and IESC
2- Moving average parameter (MA): 5 and 15
3- Safety stock level at the most downstream echelon (i.e., retai-
ler): 3 and 6
4- Safety stock level at the middle echelons (i.e., wholesaler and
distributor): 3 and 6
5- Safety stock level at the most upstream echelon (i.e., factory): 3
and 6
To conduct this experiment, we run all the possible combina-
tions of the five factors and levels, having 32 possible
Table 4
ANOVA results for the inventory variance at the different echelons.
301
combinations. Each simulation consists of 20 replications of 1200
periods each, with the first 200 periods of warm-up. In this case,
we assumed the customer demand follows a normal distribution
with lD = 30 and rD = 3. The value of rD is selected significantly
smaller than lD in order to avoid the occurrence of negative
demand observations since it is also assumed that cancelations
are not allowed which means that orders of each echelon in the
supply chain are always positive or equal to zero (Hussain et al.,
2012).
The two levels of information sharing factor are the Traditional
Supply Chain (TSC) and the Information Enriched Supply Chain
(IESC) to represent the extreme conditions of performances where
the bullwhip effect increases, respectively, geometrically and line-
arly at all echelons. The difference between TSC and IESC indicating
how the input data for the demand forecast is different in each case
is discussed in the previous section.
The experimental factor MA represents the moving average
parameter ni (i.e., ni = MA, "i) in Eqs. (13) and (14), which is the
only parameter to regulate the accuracy of the forecasting tech-
nique, being considered as one of the major causes of the bullwhip
effect in either TSC or IESC (Chen, Drezner, et al., 2000;
Dejonckheere et al., 2003, 2004). In this case, where the demand
pattern is normal, an increase in MA results uniquely in a more
accurate forecast as, according to the central limit theorem, the
estimation error decreases with the increase of the size of the sam-
ple. With this simple assumption, two levels are sufficient to
describe the relative impact that an increase of the forecast accu-
racy has on the supply chain performances whilst providing fur-
ther insights on the interaction between MA and other
experimental factors. Therefore, the values of the two levels of
MA are selected to have a high likelihood of considerable main
effect and interaction with the other parameters, whilst covering
the MA range reported in the previous literature (Chandra &
302 F. Costantino et al. / Computers & Industrial Engineering 76 (2014) 292–306

Grabis, 2005; Chatfield et al., 2004; Dejonckheere et al., 2003,
2004; Son & Sheu, 2008; Wright & Yuan, 2008). In particular, when
MA is set to the low level (ni = MA = 5, "i) then the forecast vari-
ance is r2D^ ¼ r2D =ni ¼ 1:8 that means a percentage error of about
4.5% on the average value; when MA is set to the high level (ni =
MA = 15, "i), r2D^ ¼ 0:6 and the percentage error decreases to
2.5%. If the patterns were not normal, the experimental design
should first support in finding the best value of MA, investigating
different levels of the parameter according to the weight of trend,
seasonality and autocorrelation of the demand.
In the order-up-to policy, the safety stock parameter (Ki) is the
only one to control. The work of Barlas and Gunduz (2011) showed

Fig. 11. The sensitivity of the average service level to the experimental factors.

Fig. 12. The impact of safety stock level changing on average service level.
 F. Costantino et al. / Computers & Industrial Engineering 76 (2014) 292–306
Table 5
ANOVA results for the average service level at the different echelons.
that increasing the safety stock level increases the bullwhip effect
even assuming the same Ki value at all the echelons. To deeply
investigate the interaction among the different safety stock levels
across the supply chain and identify the best allocation of inven-
tory, we define three different factors dividing the supply chain
into three zones, where the values of R_SS, WD_SS and F_SS repre-
sent the ordering level L + Ki. For instance, R_SS = 3 means that
there is no safety stock or Ki = 0 where L = 3 is the same at all ech-
elons. As for the other two parameters of the experimental design,
when the demand pattern is normal, the effect of increasing the
safety stock reflects in the same way on the supply chain perfor-
mances so two levels are sufficient to entirely describe the interac-
tions with the others.
5.1. Analysis of bullwhip effect sensitivity
Fig. 6 presents the sensitivity of the bullwhip effect of each ech-
elon i to the different factors. The figure consists of the main effects
of the factors (see, Fig. 6a–c) and of the interaction effects between
the different factors (see, Fig. 6d–f). Since the conclusions for the
bullwhip effect at the wholesaler are the same as at the distributor,
we exclude the wholesaler’s plot and present only the response for
the retailer, distributor and factory.
The strongest main effect on the bullwhip effect ratio is due to
the MA value where increasing the averaging time reduces the
bullwhip effect at all echelons. The second important factor is
the information sharing level with no impact on the retailer as it
always receives and collects the customer demand. The safety
stock level has the third important main effect where increasing
the safety stock level at any of the echelons leads to increase the
bullwhip effect at the current echelon (local effect) and at the
upstream echelons (global effect) as well. Therefore, the safety
stock policy at the downstream echelons contributes to the level
of instability at the upstream echelons. This result is a good
303
motivation for supply chain managers to seek for partnerships
with other partners in order to restrict the propagation of informa-
tion distortion. The strong interaction between MA and Info_Shar
at all echelons is evident in Fig. 7, in particular when information
are not shared, where selecting the appropriate forecasting is a
crucial task to suppress the bullwhip effect.
The other significant interactions relate to the safety stock lev-
els, confirming their role in disturbing the transmission of the
information. The two plots in Fig. 8 exhibit some examples of those
interactions at two different levels of MA.
Table 3 illustrates the analysis of variance (ANOVA) to show if
each term is statistically significant. At the retailer, where demand
is known, there is only a statistical significant effect between MA
and R_SS. At the upstream echelons there are 2-way and 3-way
interaction effects of statistical significance that confirm the find-
ings from the interaction plots.
5.2. Analysis of inventory variance ratio sensitivity
As for bullwhip effect ratio, Fig. 9 presents a sensitivity analysis
for the inventory variance ratio at the different echelons.
The main effect plots in Fig. 9a–c show that Info_Shar and MA
have the most important impact. Furthermore, the retailer’s inven-
tory performance benefits from sharing the customer demand with
the other partners in the supply chain, pushing toward collabora-
tion to reduce the propagation of the bullwhip effect whilst keep-
ing a satisfactory inventory variance in a win–win situation. The
most significant effect at the retailer is due to the WD_SS since it
is able to recover quickly from instability when the wholesaler is
holding much safety stock. Fig. 10 presents the impact of different
stock levels throughout the supply chain.
Studying the inventory variance, it is possible to confirm all the
results on the bullwhip effect, e.g. the importance of a higher MA
and information sharing. The interaction between MA and the
304 F. Costantino et al. / Computers & Industrial Engineering 76 (2014) 292–306

safety stock levels of the upper echelons influences the inventory
variance. Increasing the safety stock at echelon i leads to reduce
the inventory variance at the downstream echelon i  1 and the
decrease rate is higher if the echelon i  1 has a smaller level of
safety stock.
The ANOVA results in Table 4 show the statistical significance of
the above interactions.
5.3. Analysis of average service level sensitivity
As expected, the main effect plots in Fig. 11a–c show that the
safety stock level has the most important impact on the average
service level at all echelons. It is clear how increasing the safety
stock at any echelon i increases the service level at that echelon
and at the downstream echelons as well.
Fig. 12 details the impact of increasing the safety stock of the
retailer on the average service level at the other echelons:
increasing the R_SS level from 3.0 to 6.0 leads to increase the
average service level at the retailer from 43% to 100% while the
average service level at the wholesaler decreases from 45.3% to
40.8%, due to the amplification of the bullwhip effect and inven-
tory variance ratio.
As well, all the other factors that contribute to increase visibility
(information sharing and forecasting parameters) have a positive
effect on the service level.
Fig. 11d–f shows that the most important interaction acts
between the safety stock levels of the adjacent echelons. Increasing
the safety stock level locally (e.g., at echelon i) makes the average
service level at that echelon insensitive to whatever the safety
stock levels at the upstream echelons and especially echelon
i + 1. Larger values of safety stock at an upstream echelon improves
the average service level at the nearest downstream echelon, espe-
cially if the downstream echelon present a smaller safety stock.
Furthermore, the interaction between the safety stock levels of

Fig. 13. The ranking of the most significant terms in the ANOVA model on the bullwhip effect (a and b), inventory variance (c and d), and average service level (e and f).
 F. Costantino et al. / Computers & Industrial Engineering 76 (2014) 292–306
the distant upper echelons (e.g. i + 2 and i + 3) influences the aver-
age service level at echelon i.
The other interactions mainly show the role of safety stock in
smoothing the impact of the operational parameters on the service
level, helping to regulate the bullwhip effect and the supply chain
instability. Table 5 shows the ANOVA results for the average ser-
vice level.
6. Discussions and general implications
The scenario analysis and the experimental design provide a
clear picture of the performances that a periodic review order-
up-to level policy can generate in a supply chain. These results
present many managerial implications with a natural extension
to more complex system and real case applications. Three general
principles combine the findings, summarizing the relative main
and side effects.
6.1. Forecasting ability cannot substitute information sharing
Improving forecasting can help to reach effective results on the
bullwhip effect ratio but sharing the value of the demand has a
strongest impact on the overall performances. The ability of a sup-
ply chain partner to make accurate forecasts generates some
improvement on its local performances that can be rapidly
absorbed by the amplification of the bullwhip effect, if the other
partners do not have the same capacity. The quantification of the
inventory variance ratio and of the average service level shows that
when the collaboration level increases, the performances improve
to a great extent. The lack of visibility leads to a geometrical
increase of the bullwhip effect and to a reduction of the average
service level as moving upstream in the supply chain. The retailer,
in contact with the customer, plays a key role: as owner of the
demand information, it should be pushed to collaborate by the
smoothing of the inventory variance due to the improvement of
its supplier’s service level. The same effect can also start at any
level of the supply chain (sharing, for example, forecasts or orders
time series) but, once the bullwhip effect triggers, the possibility of
controlling performances is much limited.
6.2. Safety stock is not a synonymous of service level
The safety stock policy at the downstream echelons contributes
to the level of instability at the upstream echelons. This simple result
is as a good motivation for supply chain managers to seek for part-
nerships to restrict the propagation of information distortion.
Any coordination mechanism should provide the smallest safety
stock level at the downstream echelons to achieve a target service
level whilst protecting the upstream echelons from bullwhip prop-
agation. This propagation might also lead to increase the inventory
instability at the downstream echelons which means that there is a
negative side effect in increasing the safety stock level: only a grad-
ual adjustment of the safety stock across the supply chain (e.g.
R_SS < WD_SS < F_SS), whilst maintaining these levels as small as
possible, suppresses the bullwhip effect amplification. Just in time
policy are more effective when starting at the retailer, even without
involving all the supply chain partners.
6.3. Coordination goes beyond information sharing
All the partners should select the appropriate value of the fore-
casting parameter to achieve order smoothing and overcome the
inherent variability in incoming orders, allowing substantial
benefits for both scenarios of traditional and information enriched
supply chain. It helps to avoid the negative interactions that might
happen: as shown in Fig. 13, the moving average parameter (MA),
305
information sharing (Info_Shar) and their interactions give the
highest contribution to the bullwhip effect and the inventory var-
iance at the upstream echelon. Furthermore, these factors can act
along with the safety stock level to improve the average service
level. As information sharing level is high and moving average
parameter is accurately selected, the bullwhip effect and the
inventory variance tends to reduce and thus service level improves
without the need for much safety stock. It is clear that optimizing
supply chain is not an easy task especially in the absence of collab-
oration or information sharing. Supply chain performances not
only depend on the local operational strategies but also on the
coordination with other partners: in the above tables of ANOVA,
2-way and 3-way interactions have a statistical significance. There-
fore, without collaboration, it is very hard to gain considerable
improvements. These results provide a good motivation for initia-
tives such as VMI in which downstream echelons delegate the
upstream echelons to replenish their inventories, eliminating lay-
ers of decision making.
7. Conclusions
The lack of coordination among supply chain partners affects
their performances as a direct consequence of the bullwhip effect.
This common statement derives from the operational causes of
that generate distortion to the demand information, as for misa-
lignments of forecasting techniques, ordering policies and safety
stock levels. While many extensive researches quantified the
impact of these causes under specific conditions, agreed by all
the partners, this paper proposes a simulation study to analyze
partial levels of collaboration. In particular, the simulation study
presents the interactions that arise from different levels of infor-
mation sharing and different choices in the inventory control
parameters in a multi-echelon supply chain. The choice of the peri-
odic review order-up-to level inventory control policy supports the
extension and application of the findings to real cases where this
policy commonly applies. The results confirmed the literature on
the contribution of information sharing to the mitigation of the
bullwhip effect, revealing how also inventory variance and average
service level improves. The impact is most effective when the col-
laboration starts at the downstream echelons as, once the distor-
tion of the demand starts, it is more difficult to limit and recover.
The full factorial design examined the interactions among informa-
tion sharing and the inventory control parameters to check to what
extent they could interfere on the supply chain performances.
While confirming the role of collaboration (e.g. demand sharing)
in reducing the bullwhip effect, it was also clear how the coordina-
tion of the control policies, in term of forecasting and safety stock
level at each echelon, can support service level by reducing the
inventory variance. The general findings of the study showed that
any decision, at each echelon, has an impact locally but also trans-
mits its effect downstream and upstream, with the risk that
expected benefits reduce without coordination.
The value of the study is to represent a methodology to give
some insight about the relative contribution of the different deci-
sion leverage on the bullwhip effect and inventory stability. The
attempt to quantify the performances of a multi-echelon supply
chain with a periodic review order-up-to policy with moving aver-
age forecasting technique can be extended to evaluate the impact
of the other operational causes of the bullwhip effect. Although
many useful conclusions arose from this study, the impact of
lead-times and order batching is a main stream of evolution, as
well as other forecasting techniques. Therefore, further extension
could cover different supply chain configurations with multiple
nodes per echelon or applications of different inventory control
policies. In particular, it would be beneficial to test the interactions
of operational parameters and coordination on inventory control
306 F. Costantino et al. / Computers & Industrial Engineering 76 (2014) 292–306

policies that allow order smoothing through using control systems Disney, S. M., & Lambrecht, M. R. (2008). On replenishment rules, forecasting, and
the bullwhip effect in supply chains. Foundation and Trends in Technology,
to fit the gap between target and actual inventory and supply line
Information and Operations Management, 2(1), 1–80.
(Costantino et al., 2014b, 2014c). Following the relevant literature, Disney, S. M., & Towill, D. R. (2003a). On the bullwhip and inventory variance
this study restricts the allowance of negative customer demand produced by an ordering policy. Omega – The International Journal of
and replenishment orders as a modeling assumption. Therefore, Management Science, 31(3), 157–167.
Disney, S. M., & Towill, D. R. (2003b). The effect of vendor managed inventory (VMI)
the effect of demand variation combined with the possibility of dynamics on the Bullwhip Effect in supply chains. International Journal of
return policy (negative orders) should be studied to obtain useful Production Economics, 85(2), 199–215.
insights for supply chains with returns allowance (Chatfield & Fisher, M., Hammond, J., Obermeyer, W., & Raman, A. (1997). Configuring a supply
chain to reduce the cost of demand uncertainty. Production and Operations
Pritchard, 2013). Management, 6(3), 211–225.
Forrester, J. W. (1958). Industrial dynamics—a major breakthrough for decision
makers. Harvard Business Review, 36(4), 37–66.
Acknowledgements Holweg, M., Disney, S., Holmström, J., & Småros, J. (2005). Supply chain
collaboration: Making sense of the strategy continuum. European Management
Journal, 23(2), 170–181.
The authors would like to thank the editors and two anony- Hosoda, T., & Disney, S. M. (2006a). The governing dynamics of supply chains: The
mous reviewers for their insightful comments and suggestions on impact of altruistic behaviour. Automatica, 42(8), 1301–1309.
Hosoda, T., & Disney, S. M. (2006b). On variance amplification in a three-echelon
an earlier version of this paper. supply chain with minimum mean square error forecasting. Omega, 34(4),
344–358.
Hussain, M., Shome, A., & Lee, D. M. (2012). Impact of forecasting methods on
References variance ratio in order-up-to level policy. The International Journal of Advanced
Manufacturing Technology, 59(1–4), 413–420.
Jakšič, M., & Rusjan, B. (2008). The effect of replenishment policies on the bullwhip
Barlas, Y., & Gunduz, B. (2011). Demand forecasting and sharing strategies to reduce
effect: A transfer function approach. European Journal of Operational Research,
fluctuations and the bullwhip effect in supply chains. Journal of the Operational
184(3), 946–961.
Research Society, 62(3), 458–473.
Kelepouris, T., Miliotis, P., & Pramatari, K. (2008). The impact of replenishment
Campuzano-Bolarín, F., Mula, J., & Peidro, D. (2013). An extension to fuzzy
parameters and information sharing on the bullwhip effect: A computational
estimations and system dynamics for improving supply chains. International
study. Computers and Operations Research, 35(11), 3657–3670.
Journal of Production Research, 51(10), 3156–3166.
Klug, F. (2013). The internal bullwhip effect in car manufacturing. International
Chandra, C., & Grabis, J. (2005). Application of multi-steps forecasting for restraining
Journal of Production Research, 51(1), 303–322.
the bullwhip effect and improving inventory performance under autoregressive
Kurien, G. P., & Qureshi, M. N. (2011). Study of performance measurement practices
demand. European Journal of Operational Research, 166(2), 337–350.
in supply chain management. International Journal of Business, Management and
Chatfield, D. C. (2013). Underestimating the bullwhip effect: A simulation study of
Social Science, 2(4), 19–34.
the decomposability assumption. International Journal of Production Research,
Lau, R. S. M., Xie, J., & Zhao, X. (2008). Effects of inventory policy on supply chain
51(1), 230–244.
performance: A simulation study of critical decision parameters. Computers and
Chatfield, D. C., Hayya, J. C., & Cook, D. P. (2013). Stockout propagation and
Industrial Engineering, 55(3), 620–633.
amplification in supply chain inventory systems. International Journal of
Lee, H. L., Padmanabhan, V., & Whang, S. (1997a). The bullwhip effect in supply
Production Research, 51(5), 1491–1507.
chains. Sloan Management Review, 38(3), 93–102.
Chatfield, D. C., Kim, J. G., Harrison, T. P., & Hayya, J. C. (2004). The bullwhip effect—
Lee, H. L., Padmanabhan, V., & Whang, S. (1997b). Information distortion in a supply
impact of stochastic lead time, information quality, and information sharing: A
chain: The bullwhip effect. Management Science, 43(4), 546–558.
simulation study. Production and Operations Management, 13(4), 340–353.
Ma, Y., Wang, N., Che, A., Huang, Y., & Xu, J. (2013). The bullwhip effect on product
Chatfield, D. C., & Pritchard, A. M. (2013). Returns and the bullwhip effect.
orders and inventory: A perspective of demand forecasting techniques.
Transportation Research Part E: Logistics and Transportation Review, 49(1),
International Journal of Production Research, 51(1), 281–302.
159–175.
McCullen, P., & Towill, D. R. (2000). Practical ways of reducing bullwhip: The case of
Chen, F., Drezner, Z., Ryan, J. K., & Simchi-Levi, D. (2000). Quantifying the bullwhip
Glosuch global supply chain. IOM Control, 27(1), 24–30.
effect in a simple supply chain: The impact of forecasting, lead times, and
Nienhaus, J., Ziegenbein, A., & Schoensleben, P. (2006). How human behaviour
information. Management Science, 46(3), 436–443.
amplifies the bullwhip effect. A study based on the beer distribution game
Chen, F., Ryan, J. K., & Simchi-Levi, D. (2000). Impact of exponential smoothing
online. Production Planning and Control, 17(6), 547–557.
forecasts on the bullwhip effect. Naval Research Logistics, 47(4), 269–286.
O’Donnell, T., Maguire, L., McIvor, R., & Humphreys, P. (2006). Minimizing the
Cho, D. W., & Lee, Y. H. (2013). The value of information sharing in a supply chain
bullwhip effect in a supply chain using genetic algorithms. International Journal
with a seasonal demand process. Computers and Industrial Engineering, 65(1),
of Production Research, 44(8), 1523–1543.
97–108.
Robinson, S. (2005). Automated analysis of simulation output data. In: Proceedings
Ciancimino, E., Cannella, S., Bruccoleri, M., & Framinan, J. M. (2012). On the bullwhip
of the 2005 winter simulation conference, 1–4, 763–770.
avoidance phase: The synchronised supply chain. European Journal of
Rong, Y., Shen, Z. J. M., & Snyder, L. V. (2008). The impact of ordering behavior on
Operational Research, 221(1), 49–63.
order-quantity variability: A study of forward and reverse bullwhip effects.
Costantino, F., Di Gravio, G., Shaban, A., & Tronci, M. (2013a). Information sharing
Flexible Services and Manufacturing Journal, 20(1), 95–124.
policies based on tokens to improve supply chain performances. International
Son, J. Y., & Sheu, C. (2008). The impact of replenishment policy deviations in a
Journal of Logistics Systems and Management, 14(2), 133–160.
decentralized supply chain. International Journal of Production Economics, 113(2),
Costantino, F., Di Gravio, G., Shaban, A., & Tronci, M. (2013b). Exploring the bullwhip
785–804.
effect and inventory stability in a seasonal supply chain. International Journal of
Sterman, J. D. (1989). Modeling managerial behavior: Misperceptions of feedback in
Engineering Business Management, 5(23), 1–12. http://dx.doi.org/10.5772/56833.
a dynamic decision making experiment. Management Science, 35(3), 321–339.
Costantino, F., Di Gravio, G., Shaban, A., & Tronci, M. (2014a). Replenishment policy
Tako, A. A., & Robinson, S. (2012). The application of discrete event simulation and
based on information sharing to mitigate the severity of supply chain
system dynamics in the logistics and supply chain context. Decision Support
disruption. International Journal of Logistics Systems and Management, 18(1),
Systems, 52(4), 802–815.
3–23.
Tangen, T. (2004). Performance measurement: From philosophy to practice.
Costantino, F., Di Gravio, G., Shaban, A., & Tronci, M. (2014b). SPC-based inventory
International Journal of Productivity and Performance Management, 53(8),
control policy to improve supply chain dynamics. International Journal of
726–737.
Engineering and Technology, 6(1), 418–426.
Wright, D., & Yuan, X. (2008). Mitigating the bullwhip effect by ordering policies
Costantino, F., Di Gravio, G., Shaban, A., & Tronci, M. (2014c). Inventory control
and forecasting methods. International Journal of Production Economics, 113(2),
system based on control charts to improve supply chain performances.
587–597.
International Journal of Simulation Modelling, 13 (3), 263–275.
Xu, K., Dong, Y., & Evers, P. T. (2001). Towards better coordination of the supply
Croson, R., & Donohue, K. (2006). Behavioral causes of the bullwhip effect and the
chain. Transportation Research Part E: Logistics and Transportation Review, 37(1),
observed value of inventory information. Management Science, 52(3), 323–336.
35–54.
De La Fuente, D., & Lozano, J. (2007). Application of distributed intelligence to
Ye, F., & Wang, Z. (2013). Effects of information technology alignment and
reduce the bullwhip effect. International Journal of Production Research, 45(8),
information sharing on supply chain operational performance. Computers and
1815–1833.
Industrial Engineering, 65(3), 370–377.
Dejonckheere, J., Disney, S. M., Lambrecht, M. R., & Towill, D. R. (2003). Measuring
Zhang, X. L. (2004). The impact of forecasting methods on the bullwhip effect.
and avoiding the bullwhip effect: A control theoretic approach. European Journal
International Journal of Production Economics, 88(1), 15–27.
of Operational Research, 147(3), 567–590.
Zipkin, P. H. (2000). Foundations of inventory management. New York: McGraw-Hill.
Dejonckheere, J., Disney, S. M., Lambrecht, M. R., & Towill, D. R. (2004). The impact
Zotteri, G. (2012). An empirical investigation on causes and effects of the bullwhip
of information enrichment on the bullwhip effect in supply chains: A control
effect: Evidence from the personal care sector. International Journal of Production
engineering perspective. European Journal of Operational Research, 153(3),
Economics, 143(2), 489–498.
727–750.